Investigation of the predictability of blood glucose dynamics, based solely on preceding blood glucose measurements and implications for the treatment of diabetes mellitus

An important question about the control of blood glucose in diabetes is, Can present and future blood glucose values be predicted from recent blood glucose history? A predictive capability based only on recent blood glucose history would be advantageous because there would be no need to involve compartmental models of glucose and insulin distribution, with their inherent requirement for detailed accounting of vascular glucose loads and insulin availability. If this is possible, new, continuous blood glucose monitoring technologies under development may lead to qualitatively improved therapeutic capabilities for diabetic care. Continuous monitoring technologies may potentially alert a user to some ongoing hypoglycemia episodes or other blood glucose excursions. In this thesis, it is shown that recent glycemic measurements in conjunction with a non-compartmental model may provide sufficient information to predict near term glycemic dynamics, enabling diabetic individuals to make better treatment decisions. An analysis of published data indicates that sequences of blood glucose measurements are not random series, and that blood glucose dynamics can be predicted, at least for the near future, from sufficiently sampled blood glucose measurements. Implications of these findings are discussed and shown to be relevant to the development of glycemic monitoring devices and their roles in patient treatment.